Event-Driven Embedded Firmware for Spike Detection on a Custom MEA-based Low-Cost Neural Acquisition Platform

The use of multi-electrode arrays (MEAs) has revolutionised the study of neuronal networks in vitro, enabling the simultaneous monitoring of the electrical activity of neurons. This technology is crucial for investigating neurodegenerative diseases such as Alzheimer’s and Parkinson’s and for developing new therapeutic strategies. However, commercial MEA systems are often expensive and offer limited flexibility for custom experimental protocols. Within the framework of the NeuERA project, the eLiONS team (Department of Electronics and Telecommunications) has developed a modular and low-cost acquisition platform specifically designed to perform both amperometric and potentiometric measurements on neuronal signals recorded by an innovative 60-electrode diamond-based MEA. The core of the system is a custom-designed motherboard characterized by an event-driven architecture. Unlike traditional continuous sampling methods, the event-driven approach detects neuronal spikes only when the signals exceed predefined thresholds within specific time intervals (1ms). This significantly reduces data transmission volumes, optimizing computational resources, decreasing power consumption, and enabling efficient signal analysis. The hardware configuration features dedicated analog front-end modules for each MEA channel, where signals are amplified and band-pass filtered (150–5000 Hz) to minimize noise and artefacts. Subsequently, signals are further processed on the motherboard using instrumentation amplifiers and directed through two paths (inverted and non-inverted) to efficiently capture both positive and negative spikes. ADCs digitize these signals, enabling precise monitoring, while DACs dynamically set channel-specific thresholds. The spike detection threshold is automatically determined by the firmware, which computes it as a multiple of the standard deviation of the acquired signal. This adaptive approach ensures robust detection tailored to the characteristics of each individual channel. Schmitt Trigger comparators continuously compare each processed signal against its threshold, and OR logic gates merge the outputs of the comparators for each signal polarity, ensuring that each channel produces a single digital spike event (3.3 V or 0 V). This digital signal is read by the microcontroller as a GPIO input and triggers an event only when actual neuronal activity is present. The embedded system, based on a STM32H723ZG microcontroller, captures these spike events through Input Capture timers and communicates data to a MATLAB GUI via USB. The firmware, developed specifically for this system, manages SPI communication to configure the ADCs and DACs. It performs threshold calibration based on acquired signals and utilizes interrupt-driven routines to efficiently handle incoming spike data. The firmware implements a double-buffering strategy to prevent data loss, optimize memory management, and ensure continuous real-time operation, while concurrently formatting and transmitting structured data packets via USB to a MATLAB GUI for visualization. To thoroughly test and validate both hardware and firmware functionalities in the absence of biological samples, a specialised test board was created, incorporating an all-pass filter network to generate time-shifted spike-like signals driven by a function generator.